SFO15-411: Energy Aware Scheduling: Power vs. Performance policy Speaker: Vincent Guittot, Juri Lellli Date: September 24, 2015 ★ Session Description ★ Share the impact of various patchsets, tunables, explore ideas on what else can be done. ★ Resources ★ Video: Presentations: http://www.slideshare.net/linaroorg/sfo15411-energy-aware-scheduling-power-vs-performance-policy-part-1 http://www.slideshare.net/linaroorg/sfo15411-energy-aware-scheduling-power-vs-performance-policy-part-2 Etherpad: pad.linaro.org/p/sfo15-411 Pathable: https://sfo15.pathable.com/meetings/303083 ★ Event Details ★ Linaro Connect San Francisco 2015 - #SFO15 September 21-25, 2015 Hyatt Regency Hotel http://www.linaro.org http://connect.linaro.org

1. 1
EAS: sched-DVFS and SchedTune
Status and next steps
Juri Lelli
Patrick Bellasi

2. 2
Sched-DVFS
 Brief intro
 Current status
 Open questions/next-steps
SchedTune
 Brief intro
 Current status
 Open questions/next-steps
Can’t cover all of the above here
 Follow on deep dives in the hacking rooms
Agenda

3. 3
 Based on CFS’ per-entity load tracking (PELT)
 Plus optimizations: jump to max OPP
 Key trigger points
 enqueue_task_fair()
 dequeue_task_fair()
 task_tick_fair()
sched-DVFS: Intro

4. 4
 Mike's RFCv3 was the last posting on this topic
 In our EAS RFCv5 posting we added additional optimisations
 Want to work with Mike to plan next steps
 At LPC the landscape for sched-DVFS has changed to some extent (next slide)
 A desired outcome from this discussion:
 Who is working on what ?
 What general time scales can they commit to ?
sched-DVFS: status

5. 5
Policy implementation
 Do we have broad agreement on the current design ?
 Do we need to have a scheduler policy governor interface (similar to but different from cpufreq) ?
 Who’s working on this ?
Locking implementation
 PeterZ wants RCU locking APIs to be used in hot paths shared with cpufreq core
 Who should be working on this ?
sched-DVFS for other scheduler classes
 Reports that RT threads aren’t getting serviced (unlike interactive governor)
 Who should be working on this ?
sched-DVFS: Open questions/next-steps

6. 6
Aims to service the long standing maintainer request for a simple, central tunable for
 Power reduction vs performance boosting
Extends Sched-DVFS for OPP selection and EAS for task placement
 provides sched-DVFS with behaviours similar to other governors
e.g. Interactive, performance
 supports EAS to trade-off power reduction for performance boosting
Fosters the collection of sensible information from user-space
 To support better task scheduling decisions
 Provides a simple, effective API to middleware
eg Android, ChromeOS
SchedTune: Intro

7. 7
Posted RFCv1 on LKML [1]
 Extends sched-DVFS: biases OPP selection
 Provides global and per-task tunable (based on cgroups)
 Testing on ARM TC2,ARM Juno, partner silicon
 Used an extended version of rt-app to report a “performance index”
Has triggered an interesting discussion with lots of valuable feedback
SchedTune: status
[1] https://lkml.org/lkml/2015/9/15/679

8. 8
Evaluate the benefits of the per-task approach
 Is it worth it ?
 Is cgroups the best user space API for this ? Alternatives (syscall, prctl) ?
Integration with EAS to bias task placement
 Definition of power-performance space
 Energy_diff() filtering
 Initial prototype already available [1]
Evaluation wrt existing cpufreq governors
 What use-cases should we focus on for comparison ?
 How to setup a common workflow and shareable experimental data ?
SchedTune: Open questions/next-steps
[1] https://chromium-review.googlesource.com/#/c/296883/4